CELLeBT is a novel machine learning tool to decipher B cell-T cell interactions from single cell multi-omics data through leveraging droplets containing both a B cell receptor (BCR) and T cell receptor (TCR), thus containing both a B and a T cell. CELLeBT provides multiple types of information about B cell-T cell interactions, including the relative frequency of stable cellular interactions, the clonal nature and network of interactions, and the molecular drivers of these interactions as demonstrated through analysis of RCC and PDAC datasets, with unprecedented detail.